Real-time AI conversations: ElevenLabs Conversational AI supports WebRTC, which provides built‑in background noise removal and echo cancellation. Enable WebRTC in your client (e.g., set connectionType to "webrtc") and use the WebRTC conversation token endpoint for sessions. (elevenlabs.io)
Post‑processing/recorded audio: Use Voice Isolator (UI and API) to remove background noise from audio or video files; the API endpoint is POST /v1/audio-isolation. (elevenlabs.io)
Voice Changer: There’s a “Remove Background Noise” setting to clean your mic input when using the Voice Changer tool. (elevenlabs.io)
Voice cloning samples: When uploading/editing training samples, you can set remove_background_noise to clean the samples before training. (elevenlabs.io)
If you share your setup (web, mobile, or server), I can point you to the exact SDK/sample to enable WebRTC noise removal. (elevenlabs.io)
Short answer:
A signed URL is a temporary, authenticated WebSocket URL you generate server‑side with your ElevenLabs API key. Clients use it to start a conversation with an agent that requires authorization—without exposing your API key in the browser or app. It typically expires after 15 minutes; start the session before it expires. (elevenlabs.io)
Why you might need it:
Your agent requires authorization (enable_auth). Public agents can be started client‑side with just the agentId; private/authorized agents require a signed URL. (elevenlabs.io)
You’re building a client‑side app and want to avoid exposing your API key; signed URLs are the recommended approach. (elevenlabs.io)
You want layered security by combining signed URLs with an allowlist of permitted domains. (elevenlabs.io)
How it works (high level):
Your server calls GET /v1/convai/conversation/get_signed_url with the agent_id and xi-api-key. It returns a wss:// signed_url. (elevenlabs.io)
Your client uses that signed_url to start the Conversation session. (elevenlabs.io)
The signed URL expires after 15 minutes; existing sessions continue, but you’ll need a new signed URL to start new sessions. Generating one per user session is recommended. (elevenlabs.io)
Short answer: Use ElevenLabs Conversational AI with a WebRTC connection via the official SDKs.
What to use
Web (React): @elevenlabs/react. Start a session with connectionType: "webrtc", using either a public agentId or a conversationToken for private agents. The React docs explicitly support startSession({ agentId | conversationToken | signedUrl, connectionType: 'webrtc' }). (elevenlabs.io)
iOS (Swift): ElevenLabs Swift Conversational AI SDK supports starting sessions by agentId (and other auth flows) and handles audio tracks natively (WebRTC under the hood). (elevenlabs.io)
How to authenticate for WebRTC
For public agents: pass agentId directly to startSession with connectionType: 'webrtc'. (elevenlabs.io)
For private agents: mint a WebRTC session token on your server, then pass it to the client:
Server: GET /v1/convai/conversation/token with xi-api-key and agent_id → returns { token }. (elevenlabs.io)
Important distinction (to your earlier “signed URL” question)
Signed URLs are for WebSocket connections (you fetch a temporary wss URL). For WebRTC, use a conversation token instead—no signed URL needed. (elevenlabs.io)
Why WebRTC
Delivers built‑in echo cancellation and background noise removal and improved conversation quality compared to traditional streaming. (elevenlabs.io)
Rollout status
WebRTC support is available in the React SDK (added in v0.2.0 per docs) and used in production at ElevenLabs; Swift SDK also supports it. (elevenlabs.io)
References:
React SDK usage and WebRTC options: startSession parameters. (elevenlabs.io)
WebRTC token endpoint (server-side): GET /v1/convai/conversation/token. (elevenlabs.io)
hot solstice
ruby axle